Splitting the Indifference

Today Mic Bowman of Intel gave a talk at the Center for Computer Games and Virtual Worlds at UC Irvine in which insights about how social experiences are constructed were applied to the technical problems created by the fact that at "the core of most virtual world applications is a quadratic computation of the set of interactions among avatars (users) and a shared scene," which creates a "fundamental barrier to scaling experiences in these applications." Much of the talk was devoted to specifics of previous techniques for limiting the interaction among avatars through spatial subdivision or sharding. However, Bowman argued that these algorithmic approaches consequently caused "the content and shape of the interaction" to be "driven by the limits of the simulation." Although I might argue that all kinds of real-world social interactions are constrained by certain architectures of control, Bowman claimed that Intel's use of "distributed systems technologies to balance the simulation load dynamically and to optimize network communication" in OpenSim could create more compelling and naturalistic virtual worlds.

One of the examples that Bowman kept coming back to is the challenge of creating an experience like "watching a baseball game with your buddy" rather than just "watching a baseball game." As he pointed out, a simple mouse swipe across the screen that changes the view of the player's avatar can create many technical problems, because the associated shift in head position has be be updated from the perspective of all the other avatars in a given scene. Fortunately, he noted, there can be a lot of redundancy for purposes of rendering scenes, as in the case of an event taking place on the other side of the stadium, which probably looks very similar to the avatar sitting next to you. In a talk that detailed the challenges of partitioning for scenes and operators, accounting for a user's cone of sight, and accommodating collision detection and physics, Bowman noted that there were lessons to be learned from real-time media streaming, as well as thinking about scene management from the perspective of the client. He also observed that granularity of attention matters, since those doing protein folding on ScienceSim might have very different expectations from those planning a workshop or a business meeting.